We suggest a shading essentially based division theory using the Convolution Neural Network technique to observe tumor protests in cerebrum pictures of reverberation (MR). During this shading, the mainly based algorithmic division guideline with FCNN suggests that changing over a given dark level man picture into a shading territorial picture at that point separates the situation of tumor objects from partner man picture elective objects by fully exploiting Convolution Neural Network and bar outline package. Analysis shows that the methodology will succeed in dividing human mind images to help pathologists explicitly recognize the size and district of size.